{"ID":5937122,"CreatedAt":"2026-07-07T03:14:33.014478982Z","UpdatedAt":"2026-07-09T12:08:14.305007556Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.04978","arxiv_id":"2607.04978","title":"Qantara: Bridge-Flow Training for Multi-Paradigm JEPA Control","abstract":"Joint-Embedding Predictive Architectures (JEPAs) underpin a growing family of latent world models for control from raw pixels, but every existing JEPA world model commits at training time to a single inference paradigm: either trajectory optimisation in a learned dynamics model, or direct behaviour cloning. A single checkpoint that serves both would defer this choice to inference, when deployment constraints (rollout cost, observation accessibility) determine which path wins. We present Qantara, an end-to-end JEPA whose joint training objective pairs a Brownian-bridge interpolant between consecutive clean latents on the state axis with noise-to-data flow matching on the action axis. The same checkpoint serves three inference paradigms without retraining: latent planning, behaviour-cloning action sampling, and inverse dynamics, which we query through a video-inverse composition that first predicts the next latent without action conditioning, then extracts the action. Training concentrates mass on the edges of the (action-time, state-time) noise square, where inference queries the predictor: replacing it with uniform interior sampling drops Push-T planning from 90.1 to 53.3 SR at matched compute. On the LeWM control suite, Qantara reaches a 91.2 SR three-train-seed average and sets new SOTA on OGBench-Cube (+7.7 SR over DINO-WM, +19.7 over LeWM). From the same weights, the behaviour-cloning and video-inverse paths reach 82-83 SR on Push-T and 71-73 SR on Cube. These results move JEPA world models from single-paradigm planners to multi-paradigm controllers.","short_abstract":"Joint-Embedding Predictive Architectures (JEPAs) underpin a growing family of latent world models for control from raw pixels, but every existing JEPA world model commits at training time to a single inference paradigm: either trajectory optimisation in a learned dynamics model, or direct behaviour cloning. A single ch...","url_abs":"https://arxiv.org/abs/2607.04978","url_pdf":"https://arxiv.org/pdf/2607.04978v1","authors":"[\"Ruslan Rakhimov\",\"George Bredis\",\"Yuriy Maksyuta\",\"Daniil Gavrilov\"]","published":"2026-07-06T12:12:50Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"cs.CV\",\"cs.RO\"]","methods":"[]","has_code":false}